Output sensitivity of inflation in the euro area: indirect evidence from disaggregated consumer prices Annette Fröhling Kirsten Lommatzsch
Discussion Paper Series 1: Economic Studies No 25/2011 Discussion Papers represent the authors’ personal opinions and do not necessarily reflect the views of the Deutsche Bundesbank or its staff.
Editorial Board:
Klaus Düllmann Frank Heid Heinz Herrmann Karl-Heinz Tödter
Deutsche Bundesbank, Wilhelm-Epstein-Straße 14, 60431 Frankfurt am Main, Postfach 10 06 02, 60006 Frankfurt am Main Tel +49 69 9566-0 Telex within Germany 41227, telex from abroad 414431 Please address all orders in writing to: Deutsche Bundesbank, Press and Public Relations Division, at the above address or via fax +49 69 9566-3077 Internet http://www.bundesbank.de Reproduction permitted only if source is stated. ISBN 978-3–86558–758–9 (Printversion) ISBN 978-3–86558–759–6 (Internetversion)
Abstract: We investigate output sensitivity of inflation in the euro area through a disaggregated analysis using price indices at the COICOP 4-digit level and compare cyclical sensitivity of a newly created index of cyclically sensitive items (ICSP) with that of headline HICP and core price indices. We also relate the ICSP to the first common factor extracted from the disaggregated prices, which best reflects the common dynamics of the underlying price indices. Our results indicate that two thirds of the items in the euro area HICP are cyclically sensitive. Categories most robustly related to the business cycle are food items (processed and unprocessed), non-durable industrial goods and services related to recreation. Output sensitivity of the ICSP is significantly higher than that of headline inflation. The difference in output sensitivity is most striking between the ICSP and core inflation because of the rather strong cyclical sensitivity of processed and unprocessed food prices (both in prevalence and the estimated parameter of output sensitivity). The index of cyclically sensitive prices is highly correlated with the first common factor. Given the weak factor structure of disaggregated prices, however, we conclude that the domestic business cycle is an important determinant of inflation but it is only one among a number of nearly equally important factors.
Keywords: output sensitivity, inflation, disaggregated price indices, heterogeneity, euro area, factor analysis. JEL-Classification: E31.
Non technical summary Output sensitivity of inflation has been rather weak in the euro area, including in the past recession. This has raised the question whether the business cycle still has a role in the inflation process. Recently, it was suggested for the US that the cyclical sensitivity of individual prices might differ and that the output-inflation relationship may hold true in particular for a specific part of the consumer basket (Bryan and Meyer, 2010). The idea of the present investigation is to determine the role of the business cycle for aggregate inflation dynamics in the euro area through a test of the output sensitivity of the euro area’s aggregate and individual member states’ HICP COICOP 4-digit level price indices. Specifically, does low output sensitivity hold true for all underlying HICP components in general or does cyclical sensitivity diverge. Should it differ, can a disaggregated analysis yield new insights about the inflation process? Our research strategy consists, first, in the econometric determination of cyclical sensitivity of disaggregated consumer price indices (4-digit COICOP level of the HICP) in the euro area as a whole and in its member states. The weight of the cyclically sensitive items and the weight of these items in the broad product categories (food, industrial goods, services) are identified. Second, we create an index of cyclically sensitive prices (ICSP). Its output sensitivity can be quantified and compared with that of headline inflation and several core rates. Third, the ICSP is related to the first factor extracted from the disaggregated price indices, which is assumed to best reflect their common trend. The importance of the business cycle for aggregate inflation is thus established indirectly: if the newly created index is stronger correlated with the first factor than other price measures, this might prove that the business cycle still has an important role in the overall inflation process, not only for the prices assessed as cyclically responsive. Our results indicate that cyclical sensitivity indeed differs among the individual disaggregated price indices. In the euro area as a whole (direct estimation), a rather substantial part of the HICP (two thirds) is cyclically sensitive. However, results are heterogeneous across the member states, with the larger countries tending to show a higher share of cyclically sensitive items than the smaller member states. Output sensitivity of the ICSP is significantly higher than that of headline inflation or the index of flexible prices. However, all indices are clearly more sensitive to the business cycle than core rates excluding food and energy prices. This stems from the fact that many food prices appear to be strongly cyclically sensitive, are regards prevalence and the estimated parameter. In the euro area, the correlation of the sub-index of cyclically sensitive prices with the first common factor of the HICP components is higher than for any of the other investigated price aggregates or special indices. We take these results as evidence that the business cycle still plays an important role for inflation dynamics, in particular through specific product classes. However, the weak factor structure indicates that the business cycle is not the single decisive factor driving inflation; other factors may have nearly equal relevance. All in all, our investigation shows that the business cycle does play a role in price dynamics, although probably more so in some categories than in others.
Nichttechnische Zusammenfassung Die Reagibilität der Inflation bezüglich der Output-Lücke im Euro-Raum ist relativ gering, wie empirische Untersuchungen zeigen. Dies wirft die Frage auf, welche Rolle die Konjunktur noch im Inflationsprozess spielt. Bryan und Meyer (2010) zeigten kürzlich anhand von US-amerikanischen Daten, dass sich die zyklische Reagibilität disaggregierter Preise unterscheiden kann. Die Konjunktur spiele unverändert eine bedeutende Rolle für die Preisentwicklung, jedoch vor allem für spezifische Güterkategorien. Der vorliegende Aufsatz untersucht die Bedeutung der Konjunktur für den Inflationsprozess im Euroraum (insgesamt und für einzelne Mitgliedsländer) im Rahmen einer disaggregierten Analyse (COICOP-4-Steller des Harmonisierten Verbraucherpreisindizes, HVPI). Im Zentrum steht die Frage, ob die für den Gesamtindex beobachtete schwache Konjunkturreagibilität für alle Komponenten des Preisindex gleichermaßen gilt, oder ob sich die Konjunkturreagibilität der Preise auf disaggregierter Ebene unterscheidet. Sollte dies der Fall sein, können durch die gesonderte Betrachtung der konjunkturreagiblen Komponenten möglicherweise neue Erkenntnisse über die Dynamik der Preise gewonnen werden können. In einem ersten Schritt wird die Konjunkturreagibilität aller Teilindizes des HVPI geschätzt. Aus den reagiblen Komponenten wird in einem zweiten Schritt ein neuer Index konjunkturreagibler Preise (ICSP) berechnet. Die Output-Reagibilität dieses Index wird dann verglichen mit den entsprechenden Werten für die Gesamtinflation und für unterschiedliche Kernraten. In einem dritten Schritt wird der ICSP mit dem ersten Faktor in Verbindung gesetzt, der aus den disaggregierten Teilindizes gezogen wurde und der annahmegemäß den gemeinsamen Inflationstrend am besten abbildet. Die Bedeutung der Konjunktur für den Inflationsprozess wird somit indirekt bestimmt: wenn der neu gebildete Index reagibler Preise stärker mit dem ersten Faktor korreliert ist als andere Indizes, kann dies ein Hinweis darauf sein, dass eine Betrachtung der disaggregierten Reagibilität wichtige Informationen über den Inflationstrend liefert, die beim Blick auf den Gesamtindex oder gängige Kernraten verstellt bleiben. Unsere Untersuchungen ergeben, dass im Euro-Raum ein recht großer Teil des HVPI (zwei Drittel) konjunkturreagibel ist. Reagible und nicht-reagible Komponenten finden sich in allen großen Produktgruppen. Allerdings ist die Streuung der Ergebnisse für die Mitgliedsländer recht hoch. Die Anteile der reagiblen Teilindizes sind in den großen Mitgliedsländern in der Tendenz höher. Die Reagibilität des Index konjunkturreagibler Preise ist im Euroraum höher als die des Gesamtindex oder des Index flexibler Preise. Die Werte für alle genannten Indizes liegen aber deutlich oberhalb der Ergebnisse beispielsweise für die Kernrate ohne Nahrung und Energie. Ein Grund dafür dürfte die offensichtlich stark ausgeprägte Konjunkturreagibilität vieler Nahrungsmittel sein. Zudem ist im Euroraum insgesamt die Korrelation des ICSP mit dem ersten Faktor der Komponenten des HVPI höher als der entsprechende Wert für jeden anderen hier untersuchten Teilindex. Unsere Untersuchung bestätigt somit, dass die Preisdynamik nach wie vor auch von der Konjunktur abhängt, allerdings in einigen Kategorien stärker als in anderen. Es zeigt sich aber auch, dass weitere Faktoren für die Preisdynamik nahezu ähnlich wichtig sind.
Contents 1 Introduction
1
2 Theoretical and empirical background
4
2.1
New Keynesian Phillips Curve
2.2
Heterogeneity in price flexibility and inflation persistence and its implications for output sensitivity
3 Applied strategy and data
4
6 9
3.1
Method
3.2
HICP and components
11
3.3
Flexibility of the disaggregated price indices
13
3.4
For comparison: Core inflation and special aggregates
14
3.5
Output gap and control variables
15
4 Cyclical sensitivity of disaggregated price indices 4.1
Weight of the cyclically sensitive items in the HICP
4.2
Cyclical sensitivity of broad categories and individual items in the euro area (direct approach)
4.3
4.4
9
17 17
19
Cyclical sensitivity of broad categories and individual items in the euro area based on member states’ results (indirect approach)
22
Cyclical sensitivity and flexibility
27
5 Index of cyclically sensitive items
29
5.1
Composition of the index
29
5.2
Output sensitivity
30
6 Relation of the ICSP to the first common factor
35
7 Conclusion
38
Lists of Tables and Figures Table 4.1
Weight of the cyclically sensitive items in the euro area HICP and in selected subgroups (direct approach)
Table 4.2
Weight of the cyclically sensitive items in the euro area HICP and in selected subgroups (indirect approach)
Table 4.3
36
Correlation of selected sub-indices of the ICSP with the first factor in the euro area
Table A.1
32
Correlation of selected price indices with the first factor in the euro area and member states
Table 6.2
31
Output sensitivity of the cyclically sensitive subgroups in the euro area
Table 6.1
28
Output sensitivity of the ICSP and several other price indices in the euro area and member states
Table 5.2
26
Overlap between flexible and cyclically sensitive items in the euro area
Table 5.1
25
Items with the least robust cyclical sensitivity in the member states (HP output gap based)
Table 4.5
23
Items with the most robust cyclical sensitivity in the member states (HP output gap based)
Table 4.4
20
37
Four-digit COICOP items and broad categories in the HICP including weight, average inflation rate and indicators of cyclical sensitivity
Table A.2
Cyclical sensitivity in the euro area and member states, weighted and simple shares in the HICP
Table A.3
46
49
Cyclical sensitivity in the euro area (direct approach) according to main categories: no. of responsive items
50
Table A.4
The most and the least cyclically sensitive items (unemployment gap based)
Table A.5
Composition of the ICSP based on output gap and unemployment gap
Table A.6
53
Output sensitivity of headline inflation, the ICSP (output gap) and several core and special indices
Table A.8
52
Mean value and standard deviation of headline inflation, the ICSP and several core and special indices
Table A.7
51
54
Correlation of the price indices with the first factor (Cont. of Table 6.1)
56
Figure 3.1
Composition of the euro area HICP in 2009
12
Figure 3.2
Share of ‘flexible’ four-digit COICOP items
14
Figure 3.3
Inflation rates of selected core indices and special aggregates in the euro area
Figure 3.4
Euro area inflation and output and unemployment gap measures
Figure 4.1
28
Inflation rates of the indices of cyclically sensitive items compared with headline inflation in the euro area
Figure A.1
24
Relation between cyclically sensitive and flexible items in the euro area and the member states
Figure 5.1
18
Comparison of weights determined for output sensitivity in the member states and the euro area
Figure 4.3
16
Weight of the cyclically sensitive components in the HICP for the euro area and the member states
Figure 4.2
15
29
Output gap and headline inflation in the euro area, 19972010
57
Figure A.2
Composition of the HICP in the euro area and its member states in 2009
Figure A.3
Weight of the cyclically sensitive items including OECD measures of the business cycle
Figure A.4
60
First factor and inflation rates of several core and special indices in the euro area (direct approach)
Figure A.6
59
First factor and inflation rates of the ICSP and headline HICP in the euro area (direct approach)
Figure A.5
58
61
Share in total variance explained by the first and the first five factors in the euro area and member states
62
Output sensitivity of inflation in the euro area: Indirect evidence from disaggregated consumer prices 1
1 Introduction In the early years of the euro area, inflation appeared only weakly sensitive to the business cycle. Inflation varied only marginally and often remained slightly above the value defined as price stability by the ECB of close to but below two per cent, even in phases of weak economic performance (Altissimo et al., 2006; see Figure A.1 in the Appendix). Some authors have attributed this weak co-movement of inflation and the business cycle (“the flattening of the Phillips curve”) to the strong credibility of central banks (Ball, 2006; Roberts, 2006; Laxton and N’Diaye, 2002). Another strand of the literature has emphasised that increasing globalisation may have raised the importance of international factors for inflation dynamics (Borio and Filardo, 2007; Razin and Binyamini, 2007; White, 2008; IMF, 2006). A third explanation for the inflation inertia was that prices may not be flexible enough (“sticky”), e.g. due to adjustment costs, longer-term contracts or “sticky wages”. This nominal rigidity creates inflation persistence and prevents prices from reacting timely to changes in business conditions (Altissimo et al., 2006). In sharp contrast to the earlier part of the 2000s, inflation has fluctuated rather strongly since 2007 (see Figure A.1). In 2007 and 2008, after world-wide shocks to food and energy prices and with buoyant global growth, headline inflation in the euro area soared. During the 2008/09 recession, with marked negative output gaps, inflation fell sharply, reaching negative territory for a few months in 2009. Some authors suggested that this provides evidence of inflation dynamics still being affected by the Phillips curve relationship (Liu and Rudebusch, 2010). The link between the business cycle and inflation may still be relevant, although potentially only during periods with strongly negative output gaps (Meier, 2010; Stock and
1
Authors’ affiliation: Deutsche Bundesbank. Contact:
[email protected];
[email protected]. The authors would like to thank Sven Blank, Sandra Eickmeier, Rafael Gerke, Felix Hammermann, Christin Hartmann, Heinz Herrmann, Johannes Hoffmann, Florian Kajuth, Vladimir Kuzin, Fabio Rumler and Harald Stahl for helpful comments. Sandra Eickmeier kindly provided the codes for factor analysis. The views expressed in this paper are those of the authors and do not necessarily represent the views of the Deutsche Bundesbank or the ESCB.
1
Watson, 2010; Peach et al., 2011). However, earlier doubts about the strength of the Phillips curve relation in the euro area quickly re-emerged when inflation rates rose significantly in 2010 despite large and only slowly declining negative output gaps (European Commission, 2011). As a result, the usefulness of the output gap as an indicator of underlying price trends has remained doubtful recently. Econometric evidence of output sensitivity of inflation has also been inconclusive. Rumler and Valderrama (2008) tested the traditional and the New Keynesian Phillips curves and could establish only very modest, if any, output sensitivity of inflation in the euro area as a whole and in its member states. Mody and Ohnsorge (2007) found that inflation rates still do react to domestic influences such as the domestic output gap and labour costs if a common inflation component among the EU member states is controlled for. Also Musso et al. (2009) find evidence for the Phillips curve in the euro area even if it may have substantially flattened during the investigated time period that starts in 1970. The most recent investigations have concentrated on the link between the output gap and core inflation (excluding volatile components like food and energy), to better capture the underlying trends. Although not stated explicitly, this choice may be due to the fact that changes in the relative prices of food and energy are considered “the most common measures” of supply shocks (Ball and Mazumder, 2011:10; Gordon, 1982). European Commission staff found that “the output gap remains a significant driver of core inflation in the euro area” but “its impact is not very large” (European Commission, 2011:17). Similar research conducted at the OECD (Moccero et al., 2011) fails to establish a link between core inflation and the output gap in the euro area. The weakness of the aggregate Phillips curve relationship notwithstanding, Bryan and Meyer (2010) suggested that the cyclical sensitivity of individual prices might differ and that the output-inflation relationship may still hold true for a part of the consumer price index. At least for some prices, the output gap could still be an important determinant and a suitable indicator of price trends. With reference to the New Keynesian Phillips curve, the authors decomposed the US CPI into “flexible” and “sticky” prices based on the investigation of Bils and Klenow (2004) and showed that a price index containing the more flexible prices indeed reacts more strongly to the output gap than the price index composed of the sticky prices. We start from this suggestion and investigate output sensitivity of the euro area price indices at the COICOP 4-digit level for the time period 1997-2009. It is tested whether output sensitivity is generally weak also on a disaggregated level or whether it differs between the price indices that make up the HICP. Should it differ, how widespread and how pronounced is cyclical sensitivity at the disaggregated level? Furthermore, if cyclical responsiveness differs at
2
the disaggregated level, how does that relate to the low sensitivity observed for the aggregate? The chosen approach to aggregate output sensitivity is therefore indirect. After determining the cyclically sensitive items, the magnitude (parameter) of their combined cyclical sensitivity is estimated through a newly created index of cyclically sensitive items (ICSP) and compared with that of headline inflation and other frequently used inflation measures. While that gives an idea of the still prevailing link to the business cycle of some components of the HICP, it cannot show how important the business cycle is for aggregate inflation dynamics. We extract the cyclically sensitive items, not the cyclical component as such. The items that co-move with the business cycle can also be affected by a number of other inflation determinants. Again, to determine the importance of the business cycle for aggregate inflation dynamics, we choose an indirect approach and relate this newly created index to the first common factor extracted from the disaggregated prices. Like the ICSP, the factor is determined through extracting specific information from the underlying price indices. The first factor, however, is a purely statistical concept and reflects the strongest common dynamics in the disaggregated items. Hence, should the newly created index be strongly related to the first factor, this (indirectly) testifies to an important role of the business cycle in aggregate inflation dynamics. Before starting this investigation, two main issues have to be addressed. The first concerns the theoretical basis. It is not obvious why disaggregated prices should diverge in their cyclical sensitivity and how divergent responsiveness of disaggregated prices to the business cycle could be motivated theoretically. Admittedly, the standard NKPC establishes a clear link between individual price flexibility and marginal costs, which can be approximated by the economy-wide output gap. It is the flexible components that co-move with the business cycle. However, the NKPC assumes a common probability for price changes at the disaggregated level. It is this general probability that determines aggregate output sensitivity, not differences in flexibility among the underlying price indices. In addition, according to the hybrid NKPC, prices can be adjusted also in a backward looking manner. This weakens the clear-cut link between flexibility and cyclical responsiveness of the standard NKPC while it does not change the assumption of a common probability for prices changing. Empirically, however, it was established that price flexibility varies systematically, e.g. along product types. That observation, in turn, precludes the approximation of marginal costs for an individual firm by the aggregate output gap as in the NKPC. As a result, the relation between disaggregated prices and the output gap appears best supported by models that relate price changes to idiosyncratic and common determinants, the output gap being one of the common factors (Bryan and Cecchetti, 1993). Its importance for the individual price can then be studied empirically.
3
A second major issue arises from the fact that the definition of “domestic” influences is not devoid of ambiguities in the euro area. What is “domestic” at the euro area level need not be “domestic” for the individual country, in particular in the small member states. We therefore investigated not only the euro area aggregate but also the individual member states for robustness, while in all cases disaggregated inflation rates were related to the respective domestic output gap. However, because the importance of purely domestic cost components may depend also on the size of the country, and the euro area output gap is not a simple average of the member states’ domestic activity (in particular for the small countries it also reflects the activity in main trading partners), the relation between the average of the member states and the aggregate euro area results is not always straightforward and requires careful consideration. The article is structured as follows. In the next section, we establish the theoretical and empirical background of our investigation. Section 3 presents the data set and Section 4 the price categories and items assessed as cyclically sensitive and non-sensitive. Section 5 contains the estimates of output sensitivity for the newly created time series, ICSP, compared with other inflation rates including headline, typical core rates and an index of “flexible” prices. Subsequently, we comment on the relation of the newly created series to the first common factor extracted from the HICP sub-components and its implications for the role of the business cycle in inflation dynamics.
2 Theoretical and empirical background 2.1 New Keynesian Phillips Curve The New Keynesian Phillips Curve (NKPC) links disaggregated prices to the output gap. The model is derived from the price setting of a representative firm, which operates in an environment of monopolistic competition. The optimal price pt* is then a fix mark-up μ over marginal costs mct. It is assumed that all firms use identical technology and face an identical isoelastic demand curve (Gali, 2008). If all firms could reset prices every period, they would choose an identical price. However, only a fraction of the firms (1-θ), 0
